Experiments where we fit parameters of the model to the post-adaptation data to investigate whether the model can learn these parameters and what they would look like.

Fit 1: Fitting of Thetas and individual costs

In this model, the parameters guiding the distributions over thresholds as well as the costs can vary. There is a separate cost term for each utterance.

Norming data:

Inferred thresholds:

Model fit for norming data:

Main study data:

Inferred thresholds:

Model fit for the main study:

Fit 2: Fitting of Thetas and individual costs + fixed rat-alpha

In this model, the parameters guiding the distributions over thresholds as well as the costs can vary.

Norming data:

Inferred thresholds:

Model fit for norming data:

Main study data:

Inferred thresholds:

Model fit for the main data:

Fit 3: Fitting of Thetas and individual costs + fixed cost of might and probably

In this model, the parameters guiding the distributions over thresholds as well as the costs can vary. There is a separate cost term for each utterance. The cost of might and probably is fixed and set to 1.

Main study data:

Inferred thresholds:

Model fit for the main data:

Fit 4. Fitting in a simplified model with fixed cost-values

In this simplified model, we only used the modals from the explaining away studies, “might” and “probably”, as well as the bare and the bare-not option. The parameters guiding the distributions over thresholds could vary, but the costs were fixed.

Inferred thresholds:

Model fit for the main data:

Fit 5: Fitting in a simplified model with fixed theta values

In this simplified model, we only used the modals from the explaining away studies, “might” and “probably”, as well as the bare and the bare-not option. The parameters guiding the distributions over thresholds were fixed, but the costs could vary.

Inferred thresholds:

Model fit for the main data:

Fit 6: Fitting in a simplified model

In this simplified model, we only used the modals from the explaining away studies, “might” and “probably”, as well as the bare and the bare-not option. The parameters guiding the distributions over thresholds as well as the costs could vary.

Inferred thresholds:

Model fit for the main data: